TCM scheme with fractional bit rates, framing signals and constellation shaping

Abstract

PCT No. PCT/GB92/00562 Sec. 371 Date Dec. 1, 1993 Sec. 102(e) Date Dec. 1, 1993 PCT Filed Mar. 27, 1992 PCT Pub. No. WO92/17971 PCT Pub. Date Oct. 15, 1992.Data are transmitted using quadrature amplitude modulation to select for transmission symbols from two (or more) different signal point constellations; constellation switching being used to facilitate transmission of a non-integral average number of bits per symbol and/or frame synchronization. Trellis shaping (where redundancy is introduced and selection of constellation "regions" performed over a number of symbols so as to minimize transmitted power) is employed; switching is provided so that power information (from stores is available appropriate to the constellation in use for each respective symbol. Synchronization may alternatively be provided by substitution outside the power control loop, of a symbol from an outer region for one selected from an inner region of a constellation. In another aspect of the invention, constellation-switching is employed within frames of A symbols, where A is not a power of two.

We claim: 
     
       1. A method of transmitting data using quadrature amplitude modulation, said method comprising the steps of: assembling groups of q bits,   coding one or more of the bits of each said group by a convolutional or block code to produce an augmented group having at least q+1 bits,   selecting for each said augmented group a symbol for transmission from a signal point constellation having more than 2 q  points by using a variable mapping, the mapping being controlled by generating, for each augmented group, power signals representing the signal power corresponding to each of a plurality of alternative mappings of the group, and decoding the power signals by a Viterbi decoder to determine a mapping for that group which substantially minimizes the time averaged power of transmitted symbols,   the number of bits q per group being repetitively variable,   the signal constellation and mapping associated therewith for each group being chosen according to the value of q for that group, and   the power signals being generated in response to a signal indicative of the value of q and in response to stored information defining the mappings to produce power signals corresponding to the mappings associated with the chosen constellation.   
     
     
       2. A method as in claim 1, wherein: each constellation comprises a plurality of subgroups of points,   the mappings associated with each constellation comprise a fixed mapping between bits of the group other than said one or more bits and the points of a subgroup wherein those bits serve to select one candidate point within each subgroup and a variable mapping between the other bits of the augmented group and the subgroups so that those bits serve to select one subgroup and hence one of the candidate points.   
     
     
       3. A method as in claim 1, in which: one bit of each group is coded by a convolutional code to produce two bits, each augmented group having q+1 bits;   each constellation comprises three or four subgroups of points, each subgroup having 2 q-1  bits;   the mappings associated with each constellation comprise a fixed mapping between the q-1 bits of the group other than said one bit and the points of a subgroup, those bits serving to select one candidate point within each subgroup and a variable mapping between the other two bits of the augmented group and the subgroups so that those two bits serve to select one subgroup and hence one of the candidate points.   
     
     
       4. A method as in claim 2 in which: the value of q varies according to a predetermined pattern within a framing structure, the largest constellation having additional signal points forming an additional subgroup thereof not included in said mapping, and in which, at a predetermined position within each frame of the framing structure which employs the largest constellation, whenever one predetermined subgroup is selected for that position, a signal from the additional subgroup is transmitted in lieu thereof.   
     
     
       5. A method as in claim 1, in which: the constellation comprises a plurality of subgroups of points,   the transmission having a framing structure, and   a first variable mapping is employed for a symbol at a predetermined position within each frame of the framing structure and a second variable mapping is employed for the remaining symbols of that frame,   the first variable mapping permitting selection of a symbol from a subgroup having a larger mean power than the remaining subgroups and the second mapping not permitting selection of a symbol from that subgroup.   
     
     
       6. A method according to claim 1, in which the number of transmitted bits per symbol is a rational non-integer greater than unity which, when expressed as a ratio B/A of two integers having no common factor, the denominator A is not a power of two, and wherein each group of B bits is transmitted by A symbols, where d is an integer less than A and greater than or equal to 1, each of A-d symbols being chosen from a first signal point constellation having a first value of q and each of the d symbols being chosen from a second, larger, signal point constellation. 
     
     
       7. A method of transmitting data using quadrature amplitude modulation, said method comprising the steps of: assembling groups of q bits,   coding one or more of the bits of each said group by a convolutional or block code to produce an augmented group having at least q+1 bits,   selecting for each said augmented group a symbol for transmission from a signal point constellation having more than 2 q  points by using a variable mapping, the mapping being controlled by generating, for each augmented group, power signals representing the signal power corresponding to each of a plurality of alternative mappings of the group, and decoding the power signals by a Viterbi decoder to determine a mapping for that group which substantially minimizes the time averaged power of the transmitted symbols,   the constellation including a plurality of subgroups of points, in which the transmission has a framing structure, and   using a first variable mapping for a symbol at a predetermined position within each frame of the framing structure and a second variable mapping for the remaining symbols in that frame,   the first variable mapping permitting selection of a symbol from a subgroup having a larger mean power than the remaining subgroups and the second mapping not permitting selection of a symbol from that subgroup.   
     
     
       8. A method as in claim 7 in which the constellation has first, second, third and fourth said subgroups, each successive subgroup having progressively larger mean powers, and in which the first variable mapping does not permit selection of a symbol from the first and second subgroups. 
     
     
       9. A method as in claim 8 including determining which of the two variable mappings to use by supplying to the Viterbi decoder, for candidate points belonging to the fourth subgroup, (a) for the symbol at the predetermined position, power signals representing the true power of those points and (b) for other symbols, power signals having a value higher than the true power value to suppress selection of those candidate points. 
     
     
       10. A method as in claim 9 including the step of supplying to the Viterbi decoder for candidate points belonging to the first and second subgroups: (a) for the symbol at the predetermined position, power signals having a value higher than the true power value to suppress selection of those candidate points, and   (b) for the other symbols, power signals representing the true power of those points.   
     
     
       11. A method as in claim 10 including the step of supplying to the Viterbi decoder for a candidate point belonging to the third subgroup: (a) for the symbol at the predetermined position, power signals representing in each case a power larger than the actual power such that the mean power represented for all the points in the subgroup is substantially equal to the mean power of the points in the fourth subgroup; and   (b) for the other symbols, power signals representing the actual power of those points.   
     
     
       12. A method of transmitting data using quadrature amplitude modulation, said method comprising the steps of: assembling groups of q bits,   coding one or more of the bits of each group by a convolutional or block code to produce an augmented group having at least q+1 bits,   selecting for each said augmented group a symbol for transmission from a signal point constellation having more than 2 q  points by using a variable mapping, the mapping being controlled by generating, for each augmented group, power signals representing the signal power corresponding to each of a plurality of alternative mappings of the group, and decoding the power signals by a Viterbi decoder to determine a mapping for that group which substantially minimizes the time averaged power of transmitted symbols,   the constellation including a plurality of subgroups of points,   the symbol transmission having a framing structure,   the constellation having additional signal points forming an additional subgroup not included in said mapping, and   at a predetermined position within each frame of the framing structure, whenever one predetermined group is selected for that position, a signal from the additional subgroup is transmitted in lieu thereof.   
     
     
       13. A method of transmitting data using quadrature amplitude modulation, said method comprising the transmission of a number of bits B per frame of A symbols providing a number of bits per symbol that is a rational non-integer greater than unity which when expressed as a ratio B/A of two integers having no common factor, the denominator A is not a power of two, each group of B bits being transmitted by A-d symbols, where d is an integer less than A and greater than or equal to 1, each of the A-d symbols being chosen from a first signal point constellation and each of the d symbols being chosen from a second, larger, signal point constellation.   
     
     
       14. A method as in claim 13 wherein: one said signal constellation has a number of signal points equal to a power of two and the other said signal constellation includes a first plurality of points that is a power of two in number and a second plurality of points, that is in number half as many as the first plurality and having a higher average power than the first plurality, and   the signal points chosen from the other constellation are coded in pairs such that each pair contains at most one symbol from the second plurality of points.   
     
     
       15. An apparatus for transmitting data using a quadrature amplitude modulation, said apparatus comprising: (a) means for assembling successive groups of q bits;   (b) means for coding one or more bits of each group by a convolutional or block code to produce an augmented group having at least q+1 bits;   (c) means for selecting for each said augmented group a symbol for transmission from a signal point constellation having more than 2 q  points using a variable mapping,   (d) means to generate, for each augmented group, power signals representing the signal power corresponding to each of a plurality of alternative mappings of the group;   (e) means for receiving said power signals and, in dependence on the power signals for a plurality of groups, to determine mappings therefor which substantially minimize the time averaged power of the groups;   (f) switching means for determining a framing structure for the transmission and for selecting groups having different numbers q of bits, and different signal constellations and mappings associated therewith, to control the power signal generating means (d) to produce power signals corresponding to the selected respective mappings and constellations.   
     
     
       16. An apparatus as in claim 15, in which: each constellation comprises a plurality of subgroups of points,   the mappings associated with each constellation comprise a fixed mapping between the bits of the group other than the one or more bits and the points of a subgroup, those bits serving to select one candidate point within each subgroup and a variable mapping between the other bits of the augmented group and the subgroups, those bits serving to select one subgroup and one of the candidate points.   
     
     
       17. An apparatus for transmitting data using a quadrature amplitude modulation, said apparatus comprising: (a) means for assembling successive groups of q bits;   (b) means for coding one or more bits of each group by a convolutional or block code to produce an augmented group having at least q+1 bits;   (c) means for selecting for each said augmented group a symbol for transmission from a signal point constellation having more than 2 q  points using a variable mapping;   (d) means to generate, for each augmented group, power signals representing the signal power corresponding to each of a plurality of alternative mappings of the group;   (e) means for receiving said power signals and, in dependence on the power signals for a plurality of groups, to determine mappings therefor which substantially minimize the time averaged power of the groups;   (f) switching means for determining a framing structure for the transmission and for selecting a first variable mapping to be employed for a symbol, chosen from a constellation comprising a plurality of subgroups of points, at a predetermined position within each frame of the framing structure and a second variable mapping to be employed for other symbols chosen from the constellation, the first variable mapping permitting selection of a symbol from a subgroup having a larger mean power than the remaining subgroups and the second mapping not permitting selection of a symbol from that subgroup, to control the power signal generating means (d) to produce power signals corresponding to the selected mappings.   
     
     
       18. An apparatus for transmitting data using quadrature amplitude modulation, said apparatus comprising: (i) means for assembling groups of q bits and coding one or more of the bits of each said group by a convolutional code to produce an augmented group having at least q+1 bits;   (ii) means for selecting for each said augmented group a symbol for transmission from a signal point constellation having more than 2 q  points using a variable mapping, the mappings being controlled by generating, for each augmented group, power signals representing the signal power corresponding to each of a plurality of alternative mappings of the group, and decoding the power signals by a Viterbi decoder to determine a mapping for that group which substantially minimizes the time averaged power of transmitted symbols;   (iii) switchable means for operating said apparatus at any one of a plurality of data rates;   (iv) means for generating signals selected from a signal point constellation of a nested set of signal constellations, each larger constellation having twice as many points as, and including all points of, the next smaller constellation;   (v) each constellation including a first, second and third subgroups of points, the mappings associated with each constellation having a fixed mapping between bits of the group other than said one or more bits and the points of a subgroup, those bits serving to select one candidate point within each subgroup and a variable mapping between the other bits of the augmented group and the subgroups, those bits serving to select one subgroup and one of the candidate points;   (vi) in any constellation, the average power of the first subgroup being less than that of the other subgroups, the average power of the second subgroup being less than of the third and that of the third being less than that of any further subgroup;   (vii) the said fixed mapping in each subgroup being such that: if the points in the second subgroup are ordered in ascending order of the powers of the points having the same mapping in the first subgroup then the powers of those points are in descending order;   if the points in the third subgroup power are also ordered in ascending order of the powers of the points having the same mapping in the first subgroup then the powers of those points are in ascending order;   and if the points in a fourth subgroup (if used) also are ordered in ascending order of the powers of the points having the same mapping in the first subgroup then their powers are in ascending order;     (viii) the first subgroup of points of each larger constellation having all the points of the first and second subgroups of points of the next smaller constellation and its second subgroup including all the points of the third subgroup of the next smaller constellation;   (ix) the mapping pertaining to each signal subgroup of a larger constellation which contains two subgroups of a smaller constellation and the mapping pertaining to the latter two subgroups being related in that: the mapping of the fixedly mapped bits in the smaller subgroups is unchanged in the larger subgroup;   the mapping of an additional bit required for the larger constellation is such that, in the first subgroup, a first and second value of that bit causes selection of, respectively, points belonging to first and second subgroup of the smaller constellation and, in the second subgroup, the said first and second values of that bit cause selection of, respectively, points belonging to the fourth and third subgroup of the smaller constellation.